Merged in MON-185-add-servicebus-monitors (pull request #85)

MON-185 - Azure - ServiceBus monitors

Approved-by: Quentin Manfroi <quentin.manfroi@yahoo.fr>
Approved-by: Alexandre Gaillet <alexandre.gaillet@fr.clara.net>
Approved-by: Laurent Piroelle <laurent.piroelle@fr.clara.net>
Approved-by: Adrien Broyere <adrien.broyere@fr.clara.net>

README.md

@@ -19,6 +19,7 @@ Here is the repository organization:
         - [eventhub](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/eventhub/)
         - [iothubs](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/iothubs/)
         - [redis](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/redis/)
+        - [servicebus](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/servicebus/)
         - [sql-database](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/sql-database/)
         - [storage](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/storage/)
         - [stream-analytics](https://bitbucket.org/morea/terraform.feature.datadog/src/master/cloud/azure/stream-analytics/)

cloud/azure/README.md

@@ -21,6 +21,7 @@ Creates a set of Azure DataDog monitors for the following components :
 * Azure SQL monitors
 * Azure Redis monitors
 * Azure Event Hub monitors
+* Azure Service Bus monitors
 * Azure Stream Analytics monitors
 * Azure Storage monitors
 * Azure IOT Hub monitors
@@ -149,6 +150,10 @@ Inputs
 | redis_server_load_rate_threshold_warning | Server CPU load rate (warning threshold) | string | `70` | no |
 | redis_status_message | Custom message for Redis status monitor | string | `` | no |
 | redis_status_silenced | Groups to mute for Redis status monitor | map | `<map>` | no |
+| servicebus_status_message | Custom message for Service Bus status monitor | string | `` | no |
+| servicebus_status_silenced | Groups to mute for Service Bus status monitor | map | `<map>` | no |
+| servicebus_status_timeframe | Monitor timeframe for Service Bus status [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`] | string | `last_15m` | no |
+| servicebus_status_aggregator | Monitor timeframe aggregator for Service Bus status [available values: min, max, sum or avg] | string | `min` | no |
 | sqldatabase_cpu_message | Custom message for SQL CPU monitor | string | `` | no |
 | sqldatabase_cpu_silenced | Groups to mute for SQL CPU monitor | map | `<map>` | no |
 | sqldatabase_cpu_threshold_critical | CPU usage in percent (critical threshold) | string | `90` | no |

cloud/azure/eventhub/README.md

@@ -10,7 +10,6 @@ module "datadog-monitors-azure-eventhub" {
 
   message = "${module.datadog-message-alerting.alerting-message}"
   environment = "${var.environment}"
-  subscription_id = "${var.subscription_id}"
 }
 ```
 

cloud/azure/inputs.tf

@@ -655,6 +655,30 @@ variable "redis_server_load_rate_threshold_warning" {
   default     = 70
 }
 
+# Azure Service Bus specific variables
+variable "servicebus_status_silenced" {
+  description = "Groups to mute for Service Bus status monitor"
+  type        = "map"
+  default     = {}
+}
+
+variable "servicebus_status_message" {
+  description = "Custom message for Service Bus status monitor"
+  type        = "string"
+  default     = ""
+}
+
+variable "servicebus_status_aggregator" {
+  description = "Monitor aggregator for Service Bus status [available values: min, max, sum or avg]"
+  type        = "string"
+  default     = "min"
+}
+
+variable "servicebus_status_timeframe" {
+  description = "Monitor timeframe for Service Bus status [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`]"
+  default     = "last_15m"
+}
+
 # Azure SQL Database specific variables
 variable "sqldatabase_cpu_silenced" {
   description = "Groups to mute for SQL CPU monitor"

cloud/azure/monitors.tf

@@ -168,6 +168,22 @@ module "redis" {
   server_load_rate_threshold_warning        = "${var.redis_server_load_rate_threshold_warning}"
 }
 
+module "servicebus" {
+  source = "./servicebus"
+
+  environment = "${var.environment}"
+  message     = "${var.message}"
+  delay       = "${var.delay}"
+
+  filter_tags_use_defaults = "${var.filter_tags_use_defaults}"
+  filter_tags_custom       = "${var.filter_tags_custom}"
+
+  status_silenced             = "${var.servicebus_status_silenced}"
+  status_message              = "${var.servicebus_status_message}"
+  status_timeframe            = "${var.servicebus_status_timeframe}"
+  status_timeframe_aggregator = "${var.servicebus_status_aggregator}"
+}
+
 module "sqldatabase" {
   source = "./sql-database"
 

cloud/azure/servicebus/README.md

@@ -0,0 +1,43 @@
+Service Bus Datadog monitor
+===========================
+
+How to use this module
+----------------------
+
+```
+module "datadog-monitors-azure-servicebus" {
+  source = "git::ssh://git@bitbucket.org/morea/terraform.feature.datadog.git//cloud/azure/servicebus?ref={revision}"
+
+  message = "${module.datadog-message-alerting.alerting-message}"
+  environment = "${var.environment}"
+}
+```
+
+Purpose
+-------
+Creates a Datadog monitor with the following checks :
+
+* Service status check
+
+Inputs
+------
+
+| Name | Description | Type | Default | Required |
+|------|-------------|:----:|:-----:|:-----:|
+| delay | Delay in seconds for the metric evaluation | string | `900` | no |
+| environment | Architecture Environment | string | - | yes |
+| filter_tags_custom | Tags used for custom filtering when filter_tags_use_defaults is false | string | `*` | no |
+| filter_tags_use_defaults | Use default filter tags convention | string | `true` | no |
+| message | Message sent when an alert is triggered | string | - | yes |
+| status_message | Custom message for Service Bus status monitor | string | `` | no |
+| status_silenced | Groups to mute for Service Bus status monitor | map | `<map>` | no |
+| status_timeframe | Monitor timeframe for Service Bus status [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`] | string | `last_15m` | no |
+| status_aggregator | Monitor aggregator for Service Bus status [available values: min, max, sum or avg] | string | `min` | no |
+
+Related documentation
+---------------------
+
+DataDog documentation : [https://docs.datadoghq.com/integrations/azure/](https://docs.datadoghq.com/integrations/azure/)  
+You must search `servicebus`, there is no integration for now.
+
+Azure metrics documentation : [https://docs.microsoft.com/fr-fr/azure/monitoring-and-diagnostics/monitoring-supported-metrics#microsoftservicebusnamespaces](https://docs.microsoft.com/fr-fr/azure/monitoring-and-diagnostics/monitoring-supported-metrics#microsoftservicebusnamespaces)

cloud/azure/servicebus/inputs.tf

@@ -0,0 +1,49 @@
+# Global Terraform
+variable "environment" {
+  description = "Architecture Environment"
+  type        = "string"
+}
+
+# Global DataDog
+variable "delay" {
+  description = "Delay in seconds for the metric evaluation"
+  default     = 900
+}
+
+variable "message" {
+  description = "Message sent when an alert is triggered"
+}
+
+variable "filter_tags_use_defaults" {
+  description = "Use default filter tags convention"
+  default     = "true"
+}
+
+variable "filter_tags_custom" {
+  description = "Tags used for custom filtering when filter_tags_use_defaults is false"
+  default     = "*"
+}
+
+# Azure Service Bus specific variables
+variable "status_silenced" {
+  description = "Groups to mute for Service Bus status monitor"
+  type        = "map"
+  default     = {}
+}
+
+variable "status_message" {
+  description = "Custom message for Service Bus status monitor"
+  type        = "string"
+  default     = ""
+}
+
+variable "status_aggregator" {
+  description = "Monitor aggregator for Service Bus status [available values: min, max, sum or avg]"
+  type        = "string"
+  default     = "min"
+}
+
+variable "status_timeframe" {
+  description = "Monitor timeframe for Service Bus status [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`]"
+  default     = "last_15m"
+}

cloud/azure/servicebus/monitors-service-bus.tf

@@ -0,0 +1,34 @@
+data "template_file" "filter" {
+  template = "$${filter}"
+
+  vars {
+    filter = "${var.filter_tags_use_defaults == "true" ? format("dd_monitoring:enabled,dd_azure_servicebus:enabled,env:%s", var.environment) : "${var.filter_tags_custom}"}"
+  }
+}
+
+resource "datadog_monitor" "servicebus_status" {
+  name    = "[${var.environment}] Service Bus is down"
+  message = "${coalesce(var.status_message, var.message)}"
+
+  query = <<EOF
+      ${var.status_aggregator}(${var.status_timeframe}): (
+        ${var.status_aggregator}:azure.servicebus_namespaces.status{${data.template_file.filter.rendered}} by {resource_group,region,name}
+      ) != 1
+EOF
+
+  type = "metric alert"
+
+  silenced = "${var.status_silenced}"
+
+  notify_no_data      = true
+  evaluation_delay    = "${var.delay}"
+  renotify_interval   = 0
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = false
+  new_host_delay      = "${var.delay}"
+
+  tags = ["env:${var.environment}", "resource:servicebus", "team:azure", "provider:azure"]
+}

cloud/azure/stream-analytics/README.md

@@ -10,7 +10,6 @@ module "datadog-monitors-azure-redis" {
 
   message = "${module.datadog-message-alerting.alerting-message}"
   environment = "${var.environment}"
-  subscription_id = "${var.subscription_id}"
 }
 ```